The long-term goal of the research is to determine the extent to which psychological processes are modular in the sense that a) psychological processes can be decomposed into a set of domains and stages within domains, with each subcomponent performing different functions and manipulating different types of information, and b) the subcomponents are "blind" to the information manipulated by other domains. These general questions will be addressed in the area of language production by focussing on two problems faced by speakers in constructing a sentence: The linearization and coordination problems. The linearization problem reflects the fact that the components of a sentence must be arranged in a linear string, yet the order in which the units must be produced may differ from the order in which they are retrieved from memory. The coordination problem derives from the fact that sentence production requires the manipulation of many types of information, such as words, syntax, and rhythm, even though these sources of information may conflict at times. The proposed studies will examine how speakers solve these problems, and the answers will be related to two modular questions regarding sentence production. First, can sentence production be divided into discrete stages, with each stage performing certain tasks and handling different types of information? Second, are these stages independent in the sense that they are "blind" to the information manipulated by other stages? The first question will be addressed through a series of studies examining word order preferences in two types of sentences, with each type theoretically associated with a different stage of production. Using sentence generation and sentence memory tasks, these studies will attempt to separate two levels of sentence production and specify the types of information they each manipulate. The second question will be addressed through a series of studies designed to determine whether the coordination problem is solved by permitting different types of linguistic information to affect other types. Using sentence memory, sentence reading, and sentence generation tasks, these studies will explore, for example, whether phonological information can affect syntactic formulation, contrary to many modular predictions. A final set of studies will examine interactions between the language production system and other aspects of cognition, such as the structure of memory and language learning.